Abstract It is suggested after analysing past studies on fatigue mechanisms that the mechanism which has received most attention in the belief that it applies to metals in general is in fact a special case. The review further suggests that the mechanism does not apply to the metals of technical importance in load bearing structures, strong metals with extensive elastic ranges; and it puts forward for these metals an alternative mechanism based on a practical interpretation of their elastic range. Finally it draws attention to a practical test that might usefully complement the empirical fatigue test in assessing the stability of these technical elastic ranges.